Structural brain architectures match intrinsic functional networks and vary across domains: A study from 15000+ individuals

Na Luo, Jing Sui, Anees Abrol, Jessica A. Turner, Eswar Damaraju, Zening Fu, Lingzhong Fan, Jiayu Chen, Dongdong Lin, Chuanjun Zhuo, Yong Xu, David C. Glahn, Amanda Rodrigue, Marie T. Banich, Godfrey D. Pearlson, Vince D. CalhounbioRxiv 2019 · DOI

TL;DR

This study reveals that the physical structure of the brain closely mirrors its functional networks, with some brain regions showing stronger structural-functional alignment than others. By analyzing over 15,000 individuals, researchers created a template of these correspondences, highlighting that more "primitive" sensory areas (like vision and motor control) are more structurally consistent than complex association areas (like those involved in planning and decision-making).

What Problem Does This Paper Address?

While we know the brain is a complex network where structure and function are intertwined, the precise extent to which anatomical architecture aligns with intrinsic functional networks across different brain domains has remained an open question. Specifically, researchers wanted to understand if this structural-functional (S-F) correspondence is consistent across the entire brain or if it varies depending on the brain's functional specialization (e.g., sensory processing vs. higher-level cognition). Addressing this could lead to a more unified understanding of brain organization and potentially inform the study of neurological disorders.

Methods

The researchers employed Independent Component Analysis (ICA), a powerful technique for decomposing complex data into underlying independent sources. They applied ICA to both structural (gray matter volume) and functional (resting-state fMRI) data.

  1. Structural Decomposition: Gray matter volume data was processed using Source-Based Morphometry (SBM) via ICA to identify spatial components representing variations in gray matter across individuals.
  2. Functional Decomposition: Resting-state fMRI data was decomposed using Group ICA to identify intrinsic functional networks.
  3. S-F Correspondence: They then calculated the spatial correlation (using Pearson correlation and mutual information) between the identified structural and functional components to quantify their S-F coherence.
  4. Replication: Crucially, these analyses were performed on two large, independent datasets (a discovery cohort and the UK Biobank) to ensure the robustness and generalizability of their findings.
  5. Network Module Analysis: The identified S-F pairs were then categorized into 9 predefined functional brain networks (e.g., visual, somatomotor, default mode, frontoparietal) to compare S-F coherence across different functional domains.

Datasets

  • Discovery Dataset: 6,005 subjects with structural MRI and 7,104 subjects with resting-state fMRI data from the University of New Mexico and the University of Colorado Boulder.
  • Replication Dataset: 9,214 subjects from the UK Biobank.

The study leverages existing large-scale neuroimaging datasets and does not introduce a new benchmark dataset.

Key Insights and Ideas

  • Brain Structure Dictates Function (to a degree): The study provides strong evidence that the physical layout of the brain (gray matter distribution) is a significant determinant of its functional organization, even at the network level.
  • Domain-Specific S-F Alignment: The degree of S-F coherence is not uniform across the brain. Simpler, more evolutionarily ancient sensory and motor systems show a tighter coupling between structure and function compared to more complex, higher-order association areas.
  • The Frontoparietal Network as an Outlier: This network, heavily involved in executive functions and cognitive control, exhibits the most divergence between its structural and functional organization. This suggests greater flexibility or plasticity in these regions.
  • A Stable S-F Template: The identification of 45 replicated S-F component pairs across two large datasets provides a valuable, stable template for understanding brain organization that can be used by the broader neuroimaging community.

Key Results

  • High Correspondence in Discovery: In the discovery dataset, 62% of structural components were matched with 77.05% of functional components.
  • Replication Success: 64.28% of the identified S-F pairs were successfully replicated in the independent UK Biobank dataset.
  • Network-Specific Replication: Visual (73.33%) and Default Mode Networks (69.23%) showed high replication rates, while the Frontoparietal Network had a low replication rate (28.57%).
  • Unimodal vs. Heteromodal: Somatomotor and visual networks (unimodal) demonstrated higher S-F coherence than heteromodal association cortices, particularly the frontoparietal network.

Limitations and Future Work

  • Resolution Mismatch: A methodological limitation is the difference in resolution between structural (1mm isotropic) and functional ( >3mm) images, which can affect signal quality and interpretation. The authors addressed this by resampling structural images and focusing on gray matter.
  • Health Status Assessment: The discovery dataset lacked detailed health status information for all participants, though the replication dataset consisted of healthy subjects, mitigating this concern for the replicated findings.
  • Scanner Variability: Differences in MRI scanners and protocols between the discovery and replication cohorts could introduce variability, but the authors argue that replicating findings across these differences strengthens the generalizability of their stable S-F template.

Future work could explore how these S-F correspondences change across development, aging, and in various neurological and psychiatric conditions. The established S-F template could also be used to investigate individual differences in brain organization and their relation to cognitive abilities.

Summary updated

Abstract

Abstract Brain structural networks have been shown to consistently organize in functionally meaningful architectures covering the entire brain. However, to what extent brain structural architectures match the intrinsic functional networks in different functional domains remains under explored. In this study, based on independent component analysis, we revealed 45 pairs of structural-functional (S-F) component maps, distributing across 9 functional domains, in both a discovery cohort (n=6005) and a replication cohort (UK Biobank, n=9214), providing a well-match multimodal spatial map template for public use. Further network module analysis suggested that unimodal cortical areas (e.g. somatomotor and visual networks) indicate higher S-F coherence, while heteromodal association cortices, especially the frontoparietal network (FPN), exhibit more S-F divergence. Collectively, these results suggest that the expanding and maturing brain association cortex demonstrates a higher degree of changes compared to unimodal cortex, which may lead to higher inter-individual variability and lower S-F coherence.
BibTeX
@misc{luo2019structural,
  author = {Na Luo and Jing Sui and Anees Abrol and Jessica A. Turner and Eswar Damaraju and Zening Fu and Lingzhong Fan and Jiayu Chen and Dongdong Lin and Chuanjun Zhuo and Yong Xu and David C. Glahn and Amanda Rodrigue and Marie T. Banich and Godfrey D. Pearlson and Vince D. Calhoun},
  title = {{Structural brain architectures match intrinsic functional networks and vary across domains: A study from 15000+ individuals}},
  year = {2019},
  doi = {10.1101/2019.12.17.879502},
  howpublished = {bioRxiv},
}
Figure